Download Tutorial on Spatial and Spatio-Temporal Data Mining Part II

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Document related concepts

Nonlinear dimensionality reduction wikipedia, lookup

K-means clustering wikipedia, lookup

Cluster analysis wikipedia, lookup

Transcript
Tutorial on Spatial and Spatio-Temporal Data Mining
Part II – Trajectory Knowledge Discovery
Vania Bogorny
Universidade Federal de Santa Catarina
www.inf.ufsc.br/~vania
[email protected]
Outline
The wireless explosion
Moving Object Data and Mobility Data Analysis
Trajectory Patterns
Geometric Trajectory Pattern Mining Methods:
T3
T2
T1
Semantic Trajectory Pattern Mining Methods:
T4
C
R
R
SC
T2
H
T1
Trajectory Data mining Tools
H
HotelR
C
H
T3
T4
C
Restaurant
Cinema
The Wireless Explosion (Fosca Giannotti 2007 – www.geopkdd.eu)
Have you ever feel to be tracked?
The Wireless Explosion
The world becomes more and more mobile with the easy
access to smart phones, GPS, etc
Sattelite services, sensors and wireless technologies are
rapidly improving
lots of spatio-temporal data is being generated
A Explosão
da Rede
Sem Fio
The Wireless
Explosion
(Fosca Giannotti 2007 – www.geopkdd.eu)
Mobile devices leave behind digital traces that are collected as
trajectories, describing the movement of its users
Mobile devices generate a new type of data, called “ Trajectories of
Moving Objects”
5
Mobility Data Analysis
Mobility Data Analysis
Several analysis may be done over trajectories:
How people move around the town
During the day, during the week, etc.
Are there typical movement behaviours? In a certain area at a certain
time?
How are people movement habits changing in this area in last decadeyear-month-day?
Are there relations between movements of two areas?
Are there periodic movements?
Mobility Data Analysis: Applications
Trajectory data analysis may be useful in
several application domains
Veicule Monitoring
Transportation Companies monitor their trucks
Insurance companies use GPS devices to monitor
insured vehicles to reduce insurance price
Traffic Analysis
To alert people about traffic jams,
accidents, etc...
Identify/predict low traffic regions
in a city
Mobility Data Analysis
Animal Migration / Behaviour Analysis
Which are the trajectories of
a given migration bird?
Where do birds stop? For how long?
Which is the migration pattern of certain species?
Fishing Analysis and Control
Are boats really fishing in
allowed areas?
Can we classify vessel trajectories?
Mobility Data Analysis
Weather prediction and movement
analysis
Hurricane tracking
Trajectory Data
Trajectory Data (Giannotti 2007 – www.geopkdd.eu)
Spatio-temporal Data
Represented as a set of points, located in space and time
T=(x1,y1, t1), …, (xn, yn, tn) => position in space at time ti was
(xi,yi)
Tid
position (x,y)
1 48.890018 2.246100
1 48.890018 2.246100
...
...
1 48.890020 2.246102
1 48.888880 2.248208
1 48.885732 2.255031
...
...
1 48.858434 2.336105
1 48.853611 2.349190
...
...
2
...
time (t)
08:25
08:26
...
08:40
08:41
08:42
...
09:04
09:05
...
...
Trajectories: Overall Characteristics
1.
2.
3.
4.
(Adrienko 2008)
Geometric shape
Length (traveled distance)
Duration (in time)
Speed
Mean and maximal Speed
Acceleration, deceleration
5. Direction:
Periods of straight, curvilinear, circular movement
More.....
Relationships
Many types of relations may be of interest, depending on the problem:
similarity or difference of the overall characteristics of the trajectories
e.g. shapes, travelled distances, durations, dynamics of speed and directions)
spatial and temporal relations:
co-location in space (i.e. the trajectories consist of the same
positions or have some positions in common):
co-existence in time (i.e. the trajectories are collected during the
same time period or the periods overlap);
co-incidence in space and time (i.e. same positions are attained at
the same time);
distances in space and in time.
Trajectory Patterns
Mining Trajectories: Clustering
Fosca Giannotti 2007 – www.geopkdd.eu
Group together similar trajectories
For each group produce a summary
= cell
Mining Trajectories : Frequent patterns
Fosca Giannotti 2007 – www.geopkdd.eu
Frequent followed paths
= cell
Mining Trajectories: classification models
Fosca Giannotti 2007 – www.geopkdd.eu
Extract behaviour rules from history
Use them to predict behaviour of future users
20%
5%
7%
60%
?
8%
= cell
Trajectory Data Mining Methods
Spatio-Temporal Data Mining Methods
Two approaches:
Geometry-based spatio-temporal data mining:
Density-based clustering methods
Focus on physical similarity
Consider only geometrical properties of trajectories (space and
time)
Semantic-based spatio-temporal data mining
Deal with sparse data also
Patterns are computed based on the semantics of the data
Trajectories are pre-processed to enrich the data
Geometry-based Trajectory Data Mining
Methods
General Geometric Trajectory Patterns
Relative Motion Patterns (Laube 2004)
Proposed 5 kinks of trajectory patterns based on movement,
direction, and location: convergence, encounter, flock,
leadership, and recurrence
Convergence: At least m entities pass through the same
circular region of radius r, not necessarily at the same
time (e.g. people moving to train station)
T4
T1
T2
T3
T5
convergence
Relative Motion Patterns (Laube 2004)
Flock pattern: At least m entities are within a region of radius r and move in
the same direction during a time interval >= s (e.g. traffic jam)
Leadership: At least m entities are within a circular region of radius r, they
move in the same direction, and at least one of the entities is heading in that
direction for at least t time steps. (e.g. bird migration, traffic accident)
Encounter: At least m entities will be concurrently inside the same circular
region of radius r, assuming they move with the same speed and direction.
(e.g. traffic jam at some moment if cars keep moving in the same direction)
T2
T3
Leadership
T1
Encounter
Flock
Relative Motion Patterns (Laube 2004)
Recurrence: at least m entities visit a
circular region at least k times
F1
F1
F1
Recurrence
F1
Extension of the work proposed by [Laube 2004, 2005]
Gudmundsson(2006)
Computes the longest duration flock patterns
The longest pattern has the longest duration
And has at least a minimal number of
trajectories
Gudmundsson (2007)
proposes approximate algorithms for computing
the patterns leadership, encounter,
convergence, and flock
Focus relies on performance issues
Frequent Trajectory Patterns
Frequent Mobile Group Patterns (Hwang, 2005)
A group pattern is a set of trajectories close to each other
(with distance less than a given minDist) for a minimal
amount of time (minTime)
Direction is not considered
Frequent groups are computed with the algorithm Apriori
Group pattern: time, distance, and minsup
Co-Location Patterns (Cao 2006)
Co-location episoids in spatio-temporal data
Trajectories are spatially close in a time window and move together
w2
w1
Traclus (Han, 2007)
Clustering algorithm (TraClus-Trajectory Clustering)
Group sub-trajectories
Density-based
Partition-and-group method
1) each trajectory is partitioned into a set of line segments (subtrajectories) with lenght L defined by the user
2) similar segments (close segments) are grouped
Similarity is based on a distance function
Interesting approach for trajectories of hurricanes
Main drawback: Clustering is based on spatial distance
time is not considerd
Trajectory Sequential Patterns
Frequent Sequential Patterns (Cao, 2005)
Three main steps:
1.
Transforms each trajectory in a line with several segments
A distance tolerance measure is defined (similar to buffer)
All trajectory points inside this distance
are summarized in one segment
2. Similar segments are grouped
Similarity is based on the angle and the spatial lenght of the segment
Segments with same angle and length have
their distance checked based on a given distance d threshold
From the resultant groups, a medium segment is created
From this segment a region (buffer) is created
3. Frequent sequences of regions are computed
considering a minSup threshold
T-Patterns (Giannotti, 2007)
Sequential Trajectory Pattern Mining
Consider both space and time
Objective is to describe frequent movement
Considering visited regions of interest
During movements and the duration of movements
Steps:
1.
2.
3.
Compute or find regions of interest, based on dense spatial
regions (no time is considered)
Select trajectories that intersect two or more regions in a
sequence, annotating travel time from one region to another
Compute sequences of regions visited in same time intervals
T-Patterns (Giannotti, 2007)
Fix a set of pre-defined regions
A
B
C
Map each (x,y) of the trajectory to its region
time
Sample pattern:
A
20
min
.→ B
T-Patterns (Giannotti, 2007)
Detect significant regions thru spatial clustering
around(x1,y1)
around(x1,y1)
Map each (x,y) of the trajectory to its region
time
Sample pattern:
around ( x1 , y1 ) 20
min
. → around ( x2 , y 2 )
Trajectory Classification
The idea is to classify types of trajectories
TraClass Algorithm (Lee 2008)
Two main steps algorithm:
First: region – based clustering:
Second: trajectory-clustering
Main problem: time is not considered
TraClass Algorithm (Lee 2008)
Classify subtrajectories instead of whole trajectories
Examples:
Red trajectories move from Port A to Container Port and then to Port B
Blue trajectories move from Port A to Refinery and then to Port B
Classifying whole trajectory would classify all trajectories as moving from
Port A to Port B
38
TraClass Algorithm (Lee 2008)
First: region – based clustering
Trajectories are cut into segments (fast change of direction)
Segments are then clustered by distance with DB-SCAN
One representative trajectory is generated for the cluster and labeled with
a class
TR4
TR5
TR3
(1)A set of trajectories
(2) Partition
TR2
TR1
A representative trajectory
(3) Group
A cluster
A set of line segments
39
TraClass Algorithm (Lee 2008)
First: Discover regions that have trajectories mostly of one class regardless of
their movement patterns
TraClass Algorithm (Lee 2008)
Second: trajectory – based clustering:
Extracts clusters of common movement
patterns in non-homogeneous areas
Grouping is based on same class
41
Trajectory Outlier Detection
Trajectory Outlier Detection
•
The objective is to find trajectories that have different behavior
in relation to other trajectories
•
For instance:
–
–
–
A fishing vessel that has a behaviour different
from other fishing vessels in the same area
A hurricane that may change behaviour in certain
parts of its trajectory
Cars or pedestrians with suspishious behaviour
43
TraOD - Trajectory Outlier Detection (Lee 2008)
• Partition trajectories into subtrajectories
• Compare subtrajectories based on:
– distance and length
• If a subtrajectory is not close to other
trajectories for a minimal lenght
– It is an outlier
44
TraOD - Trajectory Outlier Detection (Lee 2008)
• Example:
– Looking to the whole trajectory, TR3 is not detected as
an outlier since its overall behavior is similar to
neigbouhr trajectories
•
Looking at the subtrajectories, T3 can be an outlier
TR5
TR
TR3 4
TR
TR1 2
An outlying sub-trajectory
45
TraOD - Trajectory Outlier Detection (Lee 2008)
Two phases: partitioning and detection
TR5
TR
TR3 4
TR
TR1 2
(1) Partition
A set of trajectories
A set of trajectory partitions
(2) Detect
TR3
An outlier
Outlying trajectory partitions
46
TraOD - Trajectory Outlier Detection (Lee 2008)
•
Once trajectories are partitioned, trajectory outliers are
detected based on both distance and density
•
A trajectory is an outlier if it contains a sufficient amount of
outlying t-partitions
Not close
> 1‒p
≤ 1‒p
Close
TRi
Li
Li is an outlying t-partition
TRi
Li
Li is not an outlying t-partition
47
TraOD - Trajectory Outlier Detection (Lee 2008)
13 Outliers from Hurricane Data
48
Summary
These data mining approaches deal with Trajectory Samples
Tid
1
1
...
1
1
1
...
1
1
...
1
1
...
1
1
1
...
2
geometry
48.890018 2.246100
48.890018 2.246100
...
48.890020 2.246102
48.888880 2.248208
48.885732 2.255031
...
48.858434 2.336105
48.853611 2.349190
...
48.853610 2.349205
48.860515 2.349018
...
48.861112 2.334167
48.861531 2.336018
48.861530 2.336020
...
...
timest
08:25
08:26
...
08:40
08:41
08:42
...
09:04
09:05
...
09:40
09:41
...
10:00
10:01
10:02
...
...
References
Laube, P. and Imfeld, S. (2002). Analyzing relative motion within groups of trackable
moving point objects. In Egenhofer, M. J. and Mark, D. M., editors, GIScience, volume
2478 of Lecture Notes in Computer Science, pages 132–144. Springer.
Laube, P., Imfeld, S., and Weibel, R. (2005a). Discovering relative motion patterns in
groups of moving point objects. International Journal of Geographical Information
Science, 19(6):639–668.
Laube, P., van Kreveld, M., and Imfeld, S. (2005b). Finding REMO: Detecting Relative
Motion Patterns in Geospatial Lifelines. Springer.
Lee, J.-G., Han, J., and Whang, K.-Y. (2007). Trajectory clustering: a partition-and-group
framework. In Chan, C. Y., Ooi, B. C., and Zhou, A., editors, SIGMOD Conference,
pages 593–604. ACM.
Li, Y., Han, J., and Yang, J. (2004). Clustering moving objects. In KDD ’04: Proceedings of
the tenth ACM SIGKDD international conference on Knowledge discovery and data
mining, pages 617–622, New York, NY, USA. ACM Press.
Nanni, M. and Pedreschi, D. (2006). Time-focused clustering of trajectories of moving
objects. Journal of Intelligent Information Systems, 27(3):267–289.
References
Verhein, F. and Chawla, S. (2006). Mining spatio-temporal association rules, sources, sinks,
stationary regions and thoroughfares in object mobility databases. In Lee, M.- L., Tan,
K.-L., and Wuwongse, V., editors, DASFAA, volume 3882 of Lecture Notes in Computer
Science, pages 187–201. Springer.
Gudmundsson, J. and van Kreveld, M. J. (2006). Computing longest duration flocks in
trajectory data. In [de By and Nittel 2006], pages 35–42.
Gudmundsson, J., van Kreveld, M. J., and Speckmann, B. (2007). Efficient detection of
patterns in 2d trajectories of moving points. GeoInformatica, 11(2):195–215.
Hwang, S.-Y., Liu, Y.-H., Chiu, J.-K., and Lim, E.-P. (2005). Mining mobile group patterns: A
trajectory-based approach. In Ho, T. B., Cheung, D. W.-L., and Liu, H., editors, PAKDD,
volume 3518 of Lecture Notes in Computer Science, pages 713–718. Springer.
Cao, H., Mamoulis, N., and Cheung, D. W. (2006). Discovery of collocation episodes in
spatiotemporal data. In ICDM, pages 823–827. IEEE Computer Society.
Zhenhui Li, Jae-Gil Lee, Xiaolei Li, Jiawei Han: Incremental Clustering for Trajectories.
DASFAA (2) 2010: 32-46
More...
Huiping Cao, Nikos Mamoulis, David W. Cheung: Discovery of Periodic Patterns in Spatiotemporal
Sequences. IEEE Trans. Knowl. Data Eng. 19(4): 453-467 (2007)
Panos Kalnis, Nikos Mamoulis, Spiridon Bakiras: On Discovering Moving Clusters in Spatiotemporal Data. SSTD, 364-381 (2005)
Florian Verhein, Sanjay Chawla: Mining spatio-temporal patterns in object mobility databases.
Data Min. Knowl. Discov. 16(1): 5-38 (2008)
Florian Verhein, Sanjay Chawla: Mining Spatio-temporal Association Rules, Sources, Sinks,
Stationary Regions and Thoroughfares in Object Mobility Databases. DASFAA, 187-201
(2006)
Cao, H., Mamoulis, N., and Cheung, D. W. (2005). Mining frequent spatio-temporal sequential
patterns. In ICDM ’05: Proceedings of the Fifth IEEE International Conference on Data Mining,
pages 82–89, Washington, DC, USA. IEEE Computer Society.
Jae-Gil Lee, Jiawei Han, Xiaolei Li, and Hector Gonzalez, “TraClass: Trajectory Classification
Using Hierarchical Region-Based and Trajectory-Based Clustering”, Proc. 2008 Int. Conf.
on Very Large Data Base (VLDB'08), Auckland, New Zealand, Aug. 2008.
Jae-Gil Lee, Jiawei Han, and Xiaolei Li, "Trajectory Outlier Detection: A Partition-and-Detect
Framework", Proc. 2008 Int. Conf. on Data Engineering (ICDE'08), Cancun, Mexico, April
2008.
Semantic-based Spatio-temporal Data Mining
Methods
Semantic Trajectory Data Mining
The main idea is to enrich trajectories with domain
semantic information in preprocessing steps
This task can be done using data mining
Apply data mining as a second step
Mining is on semantic rich trajectories
Geometric Patterns X Semantic Patterns (Bogorny 2008)
Geometric Pattern
TP
R
TP
R
CC
T2
CC
T3
T3
T2
T1
T4
H
T1
H Hotel
H
T4
R
Restaurant TP Touristic
Place
Semantic trajectory Pattern
(a) Hotel to Restaurant, passing by CC
(b) go to Cinema, passing by CC
Geometric Patterns X Semantic Patterns (Bogorny 2008)
There is very little or no semantics in most DM approaches for
trajectories
Consequence:
• Patterns are purely geometrical
• Difficult to interpret from the user’s point of view
• Do not discover semantic patterns,
which can be independent of spatial location
DJ-Cluster (Zhou 2007)
DJ-Cluster is a variation of DBSCAN
Focus relies on performance issues
Objective: find interesting places of individual trajectories
Clusters are computed from a SET of trajectories of the same object
Time is not considered
A Conceptual View on Trajectories (Spaccapietra 2008)
A trajectory is a spatio-temporal thing (an object) that
has generic features
generic: application independent
has semantic features
semantic: application dependent
A trajectory is more than a moving object
Semantic Trajectories - Motivation
Trajectory Samples (x,y,t)
Geographic Data
Geographic Data +
Trajectory Data =
Semantic Trajectories
The Model of Stops and Moves (Spaccapietra 2008)
STOPS
Important parts of trajectories
Where the moving object has stayed for a
minimal amount of time
Stops are application dependent
Tourism application
– Hotels, touristic places, airport, J
Traffic Management Application
– Traffic lights, roundabouts, big eventsJ
MOVES
Are the parts that are not stops
Semantic Trajectories
A semantic trajectory is a set of stops and moves
Stops have a place, a start time and an end time
Moves are characterized by two consecutive stops
Methods for Adding Semantics to Trajectories
Pre-processing Single Trajectories
Methods to Compute Stops and Moves
1) IB-SMoT (INTERSECTION-based)
Interesting for applications like tourism and urban planning
2) CB-SMoT (SPEED-based clustering)
Interesting for applications where the speed is important,
like traffic management
3) DB-SMOT (DIRECTION-based clustering)
Interesting in application where the direction variation is important
like fishing activities
IB-SMoT
(Alvares 2007a)
A candidate stop C is a tuple (RC, ∆C), where
RC is the geometry of the candidate stop (spatial feature type)
∆C is the minimal time duration
E.g. [Hotel - 3 hours]
An application A is a finite set
A = {C1 = (RC1 , ∆C1 ), …, CN = (RCN , ∆CN)} of candidate
stops with non-overlapping geometries RC1, … ,RCN
E.g. [Hotel - 3 hours, Museum – 1 hour]
IB-SMoT
(Alvares 2007a)
A stop of a trajectory T is a place that is important for the
application
A move of T with respect to an application is:
a maximal contiguous subtrajectory of T :
between the starting point of T and the first stop of T; OR
S1
between two consecutive stops of T; OR
S2
between the last stop of T and the ending point of T;
S3
or the trajectory T itself, if T has no stops.
IB-SMoT
(Alvares 2007ª)
Input: candidate stops
// Application
trajectories
// trajectory samples
Output: Semantic rich trajectories
Method:
For each trajectory
Check if it intersects a candidat stop for a minimal amount of time
Jurere
09-12
IbisH.
13-14
FloripaS
16-17
CB-SMoT: Speed-based clustering
(Palma 2008)
• Clusters single trajectories based on the speed variation:
low speed important place
CB-SMoT: Speed-based clustering
Unknown stop
Jurere
09-12
(Palma 2008)
Input: Trajectory samples
Speed variation
minTime
Output: stops and moves
Step 1: find clusters
Step 2: Add semantics to each
cluster
2.1: If intersects α during ∆tα stop α
IbisH.
13-14
FloripaS
16-17
2.2: If no intersection
during ∆t unknown stop
CB-SMoT: Speed-based clustering
(Palma 2008)
Unknown Stops (CB-SMOT)
same unknown stop
T1
T2
another unknown stop
CB-SMoT: Speed-based clustering
(Palma 2008)
Can Find Clusters Inside Buildings
p1
p6
p7
p11
t6= 10:10AM
t7= 10:32AM
DB-SMOT : Direction-based Clustering (Manso 2010)
Input:
trajectories
minDirVariation
minTime
maxTolerance
// trajectory samples
// minimal direction variation
// minimum time
Output: semantic rich trajectories
Method:
For each trajectory
Find clusters with direction variation
higher than minDirVariation
For a minimal amount of time
Examples of semantic trajectory patterns
Semantic Rich Trajectories (Transportation Application)
IB-SMoT
CB-SMoT
Sequential Patterns (Transportation Application)
Fishing Domain
DB-SMoT Method
Multiple-granularity semantic trajectory
pattern mining
STOPS at Multiple-Granularities (Bogorny 2009)
Stop at Ibis Hotel from 6:04PM to 7:42PM, september 16, 2010
space
time
IbisHotel or Hotel or Accommodation
Afternoon or Thursday or 6:00PM – 8:00PM or RUSH-HOUR
ITEMS - the building blocks for semantic pattern discovery
An item is generated either from a stop or a move
An item is a set of complex information (space +
time), that can be defined in many formats/types
and at different granularities
Building an ITEM for Data Mining (Bogorny 2009)
Formats/types for an item:
>ameOnly: is the name of the stop/move
STOPS: name of the spatial feature instance
• IbisHotel
MOVES: name of the two stops which define the move
• SydneyAirport – IbisHotel
.ameStart: is the name of the stop/move + start time
IbisHotel [morning]
--stop
LouvreMuseum [weekend]
--stop
IbisHotel-SydneyAirport [10:00AM-11:00AM] --move
10/11/2010
GIScience 2010 – A conceptual data model for trajectory data mining
Vania Bogorny, Universidade Federal de Santa Catarina, Brazil, www.inf.ufsc.br/~vania
79
Building an ITEM for Data Mining (Bogorny 2009)
>ameEnd: name of a stop/move + end time
IbisHotel[morning]
stop
IbisHotel-SydneyAirport[10:00AM-11:00AM]
move
>ameStartEnd: name of a stop/move + start time + end time
IbisHotel[08:00AM-11:00AM][1:00pm-6:00pm] stop
LouvreMuseum[morning][afternoon]
stop
SydenyAirport– IbisHotel [10:00AM-11:00PM] [10:00AM-6:00PM]
10/11/2010
GIScience 2010 – A conceptual data model for trajectory data mining
Vania Bogorny, Universidade Federal de Santa Catarina, Brazil, www.inf.ufsc.br/~vania
80
Multiple-Granularity Semantic Trajectory DMQL (Bogorny 2009)
ST-DMQL is an approach to semantically enrich
trajectories with domain information
Autormatically tranforms these semantic information into
different space and time granularities
Extracts frequent patterns, association rules and
sequential patterns from semantic trajectories
Multiple Level Semantic Sequential Patterns
Large Sequences of Length 2 (ITEM=SPACE+Start_Time)
(41803_street_5, 41803_street_5)
Support: 7
(41803_street_4, 41803_street_4)
Support: 9
(41803_street_4, 66655_street_4)
Support: 5
(41803_street_2, 41803_street_2)
Support: 6
(41803_street_8, 41803_street_8)
Support: 5
(41803_street_3, 0_unknown_3)
Support: 5
time unit = month
gid
Spatial feature type (stop name)
Multiple Level Semantic Sequential Patterns
Large Sequences of Length 2 (ITEM=SPACE+Start_Time)
(41803_street_tuesday,41803_street_tuesday) Support: 9
(41803_street_tuesday,66655_street_tuesday) Support: 5
(41803_street_monday,66655_street_monday) Support: 5
(41803_street_monday,41803_street_monday) Support: 11
(41803_street_monday,0_unknown_monday)
Support: 5
(41803_street_thursday,41803_street_thursday) Support: 13
(41803_street_thursday,0_unknown_thursday) Support: 6
(41803_street_wednesday,41803_street_wednesday)
Support: 7
Time unit = Day of the week
gid
Spatial feature type (stop name)
WEKA-STPM
The previous semantic pattern mining approaches are
implemented in WEKA
Weka-STDM
Weka-STDM
Current Works on Weka-STPM
Trajectory Visualization
Trajectory Cleaning
New methods for trajectory pre-processing
Trajectory Behaviour Patterns
Recent works have emerged on mining behaviour
patterns from trajectories
Athena (Baglioni 2009)
Semantic-rich movement analysis
Which are the homework trajectories? And
the common behaviors
of them?
To answer these questions
we need to define what is a
home-work trajectory (or
pattern)
The concept of the homework trajectory can be
encoded in a formal
framework to automatically
infer which trajectories are
home-work
Athena (Baglioni 2009)
Supports the post processing / deductive phase of
the KDD process
Based on ontologies to represent domain knowledge
and to infer the semantic types of the
patterns/trajectories.
Semantic classification of patters/trajectories in
domain concepts based on the semantic
characteristics
Athena (Baglioni 2009)
1.
Example
Stop
Stop
Move
Move
Semantic
Semantic
Trajectory
Trajectory
Commuter
Commuter
Trajectory
Trajectory
ontology
Commuter trajectory≡ a trajectory frequently
starting outside the city, stopping inside the city for a long
time and going back outside the city
Athena (Baglioni 2009)
SELECT t.id, t.object
FROM Milano_tr
WHERE ‘Commuter’ in
SEMANTIC(t.object)
Given the trajectories, we query the system to
identify the ones whose type is commuter, i.e.
satisfying the ontology definition
Pattern Interpretation (Rebeca 2010)
This work focuses on postprocessing, trying to interpret the
patterns
Considering that the movement
context is essential to correctly
interpret and understand the
patterns
CONTEXT = geography + thematic
attributes
Pattern Interpretation (Rebeca 2010)
1.
2.
3.
Mining movement patterns (stops)
Semantic enrichment: annotates patterns with
information obtained from trajectory types defined in
ONTOLOGIES (e.g. the trajectory of a commuter)
Use the enriched representation to automatically
classify patterns
Works Summarized in this Tutorial
Geometric Pattern
Mining Methods
(mining is on sample
points)
Laube 2004, 2005
Hwang 2005
Gudmundson 2006, 2007
Giannotti 2007
Lee 2007
Cao 2006, 2007
Lee 2007, 2008a, 2008b
Li 2010
Semantic Pattern Mining
Behaviour Pattern
Methods (Generate
Mining and
Semantic Trajectories using Interpretation Methods
DM - mining is on Semantic
Trajectories)
Alvares 2007
Zhou 2007
Palma 2008
Bogorny 2009
Bogorny 2010
Manso 2010
Alvares 2010
Giannotti 2009
Baglioni 2009
Rebeca 2010
References
Bogorny, V. ; Bart Kuijpers, Luis Otávio Alvares: ST-DMQL: A Semantic Trajectory
Data Mining Query Language. International Journal of Geographical Information
Science 23(10): 1245-1276 (2009)
Palma, A. T; Bogorny, V.; Kuijpers, B.; Alvares, L.O. A Clustering-based Approach
for Discovering Interesting Places in Trajectories. In: 23rd Annual Symposium on
Applied Computing, (ACM-SAC'08), Fortaleza, Ceara, 16-20 March (2008) Brazil.
pp. 863-868.
Spaccapietra, S., Parent, C., Damiani, M. L., de Macedo, J. A., Porto, F., and
Vangenot, C. (2008). A conceptual view on trajectories. Data and Knowledge
Engineering, 65(1):126–146.
Alvares, L. O., Bogorny, V., Kuijpers, B., de Macedo, J. A. F., Moelans, B., and
Vaisman, A. (2007b). A model for enriching trajectories with semantic geographical
information. In ACM-GIS, pages 162–169, New York, NY, USA. ACM Press.
Rebecca Ong, Monica Wachowicz, Mirco 4anni, Chiara Renso, From Pattern
Discovery to Pattern Interpretation in Movement Data - IEEE SADM 2010
References
Manso, J. A. ; TIMES, V. C. ; Oliveira, G. ; ALVARES, L. O. ; BOGORNY, V. . DBSMoT: A Direction-Based Spatio-Temporal Clustering Method. In: IEEE
International Conference on Intelligent Systems (IS), 2010, Londres. Proceedings of
the IEEE International Conference on Intelligent Systems, 2010. p. 114-119. 3.
Alvares, Luis O. ; PALMA, Andrey ; Oliveira, G. ; BOGORNY, V. . Weka-STPM:
from trajectory samples to semantic trajectories. In: Workshop de Sofware Livre,
2010, Porto Alegre. WSL, 2010.
Zhou, C.; Nupur Bhatnagar, Shashi Shekhar, Loren G. Terveen: Mining Personally
Important Places from GPS Tracks. ICDE Workshops 2007: 517-526
Bogorny, V. ; Carlos Alberto Heuser, Luis Otávio Alvares: A Conceptual Data Model
for Trajectory Data Mining. GIScience 2010: 1-15
Miriam Baglioni, José Antônio Fernandes de Macêdo, Chiara Renso, Roberto
Trasarti, Monica Wachowicz: Towards Semantic Interpretation of Movement
Behavior. AGILE Conf. 2009: 271-288
Fosca Giannotti, Mirco Nanni, Dino Pedreschi, Chiara Renso, Roberto
Trasarti: Mining Mobility Behavior from Trajectory Data. CSE (4) 2009:
948-951
Summary, Challenges and Open Issues in SpatioTemporal Data Mining
Challenges and Open Issues in Spatio-Temporal Data Mining
Trajectory Clustering
Most works are density-based clustering methods
Most are adapted spatial or non-spatial clustering algorithms
Consider either time or space, only a few consider both dimensions
Challenges and Open Issues in Spatio-Temporal Data Mining
Trajectory Similarity
Focus relies on objective similarity measures
Shape, direction, closeness
Needs: semantic similarity
Higher abstraction level similarity
Example:
– groups of trajectories going together for shopping
– Groups of trajectories going together to the University two times
a week
Challenges and Open Issues in Spatio-Temporal Data Mining
Need for data mining methods using:
Metadata
Domain knowledge
Semantics
Ontologies
For:
Trajectory data pre-processing
Pattern pruning
Improve the quality of the patterns
Pattern interpretation
More needs
There is a need for collaboration between data miners and
domain experts (environmental experts, transportation
managers, metheorologists, etc)
to evaluate data mining methods and the discovered patterns
Post-Processing: almost no spatial or spatio-temporal data
mining methods evaluate the patterns and their
interestingness
Thank You !
[email protected]